Estimation of Direct and Diffuse Solar Radiation Components from Global Solar Radiation Using CNN-LSTM Hybrid Model

Abstract

This study presents a hybrid deep learning approach utilizing Convolutional Neural Networks (CNN) and Long Short-Term Memory networks (LSTM) to estimate Direct Normal Radiation (DNR) and Diffuse Solar Radiation (DSR) from Global Solar Radiation (GSR). Leveraging temporal patterns and spatial dependencies in solar radiation data, the proposed model aims to provide accurate predictions of solar components critical for solar energy systems. Our model is trained and validated using a real-world dataset comprising daily solar radiation measurements. The CNN-LSTM model outperforms traditional machine learning methods in both accuracy and robustness.